Allocating the proper resources for your job is sometimes more of an art than a science. Start small, troubleshoot, then scale up. If you have questions please email lab AT stat.wisc.edu. We are happy to consult with you on your specific job configuration. If your job is running but inefficiently, you may hear from us as we try to fine-tune your job to achieve maximum throughput. Abusive behavior or methods that consume too many resources will not be tolerated and will be canceled by an administrator. In a multi-user environment such as an HPC cluster, it is not wise to assume more is always better. There are often diminishing returns to using as many resources as possible. Is using another 24 cores for your job to save 20 minutes of compute time worth it when others need to use the resources as well? Strive for a good balance.
Refer to the man pages
The man pages for the slurm commands are available on the SLURM Website. The most important reference is probably the man page for the
sbatch command found here. There are many, many options to set in your submit script in order to configure the best possible allocation for your work.